Powerful Open Source AI Models Being Given Away Is Sign of the Market Working

Tech ALWAYS devours its own. One of the reasons why the stock market is wavering this summer is concern about spending on AI, as investors begin to question whether those costs can ever be recovered, let alone generate the promised excess returns. 

And a major cause of that growing skepticism is the release of a wave of open source AI models, many from China, which deliver comparable performance at a significantly lower price. For those who may have forgotten Econ 1 (not that anyone blames you...) that is what a functioning market does. While American tech firms are scurrying to Washington and asking any politician who will listen (while taking donations) to outlaw China, cuz...unfair, un-American, oh, and. by the way, they're Chinese...the reality is that a number of US AI companies including - GASP - Nvidia, are releasing their own open source models because they recognize that the underlying premise - a bigger market means more sales of everything (especially more chips) - is good for business. So Silicon Valley can bitch and point fingers all it wants, but its own history teaches that this is where the market is going, so get on board or get left behind with a nice photo of your shuttered data center. JL

Bill Gurley reports in the Washington Post:

Anthropic and OpenAI are preparing to go public at valuations near $1 trillion - based on extraordinary profit margins. The challenge to those margins is a wave of powerful AI models being given away for free. Releasing an AI model openly is a business strategy with a three-decade record of building large, profitable companies. The challengers won by delivering a comparable product at a lower cost. For software, once the first copy exists, the millionth costs zero to make. A model you can download is a model you can own freeing customers from a vendor’s pricing, keeping data secure. China’s open models come from DeepSeek, whose first round valued it at $50 billion; Moonshot, which raised $2 billion; and Zhipu which is publicly traded in Hong Kong at a market cap of $100 billion. Last week, Thinking Machines released an open model. Nvidia has released its Nemotron models freely, betting it sells more of the chips underneath.  The open-model wave is not an attack on AI companies. It is the market responding to the fortune they are about to make.

Two of the most valuable companies in American artificial intelligence, Anthropic and OpenAI, are preparing to go public at valuations near $1 trillion — valuations that rest on the promise of extraordinary profit margins. The fastest-growing challenge to those margins is not another well-funded AI lab. It is a wave of powerful AI models being given away for free. The newest of them was released in China just last week. Lobbyists are urging Washington to treat open-model AI as a security threat. In fact, it is something more familiar: proper competition that should be welcomed. 

Releasing an AI model openly is not a sign of tech idealism or of a foreign company’s gambit. Open-sourcing is a business strategy, one with a three-decade record of building large, profitable companies. Treating open-model AI instead as a danger to be licensed and contained is a category error, and would be a costly one. It mistakes the ordinary work of competition for a threat to be policed. 

Sears was not allowed to outlaw Walmart. IBM was not allowed to outlaw Dell. The legacy airlines were not allowed to outlaw Southwest (though they tried). In each case, a comfortable incumbent met a rival with a lower cost structure, and in each case the incumbent had exactly two options: compete, or lose.

There is a name for what those challengers did. Economist Michael E. Porter called it cost leadership and placed it among a handful of fundamental business strategies. Cost leadership, he argued, is the surest route to a market position a rival cannot easily attack. Walmart, Dell, Southwest — none of them won by building the fanciest product. They won by delivering a comparable one at a structurally lower cost, and letting price do the rest. Open-source software is the purest form of software cost leadership ever devised.

The classical economists Adam Smith and David Ricardo would not have found any of this surprising. In a competitive market, they observed, price is driven toward the cost of producing one more unit. For most goods, such as steel, wheat or labor, that floor is stubbornly high. For software, it is almost nothing: Once the first copy exists, the millionth costs essentially zero to make. A price that falls toward zero is therefore not a market failing. It is a market working exactly as the classical economists said it would. 

The puzzle worth explaining is not why open models are free. It is why proprietary software has commanded enormous margins since the 1980s, with an artificial scarcity, fenced off by copyright and lock-in, holding price far above the cost of production. Open source simply removes the fence. And someone still makes it work, because they can profit from selling what sits next to the free thing.

The evidence is 30 years old. Red Hat built a business on freely available software and in 2019 sold to IBM for $34 billion. Confluent, MongoDB, HashiCorp, Databricks — all are large, successful companies built on code anyone can download. Google gave away Android and Kubernetes; Microsoft, which once called open source a cancer, now owns the open-source platform GitHub. None of them were confused about how to make money.

Open models are also this year’s news. China’s most capable open models come not only from its state giants, but also from independent labs — DeepSeek, Moonshot, Zhipu — and all three are backed by serious money. DeepSeek’s first outside round valued it above $50 billion. Moonshot has raised $2 billion. Zhipu is already publicly traded in Hong Kong, where investors have handed a company that gives its models away a market capitalization north of $100 billion. Whatever else one thinks of these labs, the notion that openness has no business model is not a claim their balance sheets support. 

On Friday at the 2026 World Artificial Intelligence Conference in Shanghai, Chinese President Xi Jinping urged countries to “encourage open source, openness, collaboration and sharing.” Some have tried to suggest that China might back away from open models, but in fact Xi doubled down.

Open AI models are also emerging at home. Reflection, founded by veterans of Google DeepMind, has raised billions on an explicit promise to build an open American frontier. Last week, Thinking Machines — the lab founded by OpenAI’s former chief technology officer — released a capable open model of its own. Even Nvidia has joined in, releasing its Nemotron models freely, betting that a cheap and abundant model layer sells more of the chips underneath. The open frontier is no longer just a Chinese story. It is an American one too.

A market growing fast, with rich potential margins and a long runway, is the most powerful lure in capitalism — a signal, broadcast to every capable engineer and every investor on the planet, that there is enormous money to be made by offering the same thing for less. Jeff Bezos (Amazon’s executive chairman and owner of The Washington Post) built a trillion-dollar company on the principle that “your margin is my opportunity.” By that logic, a business promising the highest returns in software summons the largest crowd of competitors in software. The open-model wave is not an attack on AI companies. It is the market responding to the fortune they say they are about to make. They called it forth themselves. 

And it was never only about price. A model you can download is a model you can own, and ownership throws off advantage after advantage for its users. It prevents lock-in, freeing you from any single vendor’s pricing, road map or survival. It keeps your data and your methods behind your own walls, rather than feeding them into — and having them absorbed by — someone else’s model. It lets anyone adapt, fine-tune or extend the system without asking permission. It compounds innovation, because thousands of independent hands improve it in parallel, faster than any one company’s road map allows. 

And, counterintuitively, it is more secure: A model anyone can inspect becomes more secure with every set of eyes on it — stress-tested, audited and hardened in the open. A generation of open-source software has proved the point: The most trusted code on earth is open precisely because a million adversarial eyes find and fix what any closed team would miss. The companies now calling open models a hazard rely on that very code every day.

Add it up, and a striking pattern emerges. Start-ups want open models, to build without a gatekeeper. Enterprises want them, because they don’t want their intellectual property being sucked into someone else’s AI model. Researchers, universities and security teams want them, to see inside what they study and defend. Chipmakers want them, because a cheap and abundant model layer sells more chips. Cloud providers without a model of their own want them. Foreign governments and militaries want them, to run systems no other nation can switch off. 

Nearly everyone in the AI economy has a reason to prefer an open foundation — everyone, that is, except the big incumbents Anthropic and OpenAI, whose fortunes depend on keeping it closed. This is not a fringe of the industry against the mainstream. It is almost the entire field on one side, and a handful of very large companies on the other.

There is a catch, and it is why this fight matters. Historically, software markets do not reward the many; they tip toward the few. Economists call it increasing returns — some know it as network effects — and economist W. Brian Arthur put it plainly: In these industries, there is a “tendency for that which is ahead to get further ahead.” AI may intensify this loop like nothing before it: A company that learns from hundreds of millions of customers, and now claims its models are helping build their own successors, compounds a lead that rivals cannot catch. This is how three firms — Microsoft, Google, Meta — came to own the internet’s core layers; Anthropic and OpenAI are on exactly that path. Left alone, this market tips. The one force that keeps it open, that keeps the many in the game against the few, is a freely available model no one can corner or shut down. 

Which brings us to Washington. Lobbyists are urging lawmakers to treat open models as a danger — to regulate them, restrict them, and wall off the foreign ones. But recall what these same large companies say about themselves: the fastest growth in history, the richest margins in software, models accelerating away from all rivals. Grant every word of it. Then ask the obvious question: If all of that is true, why do you need the government’s help? A company running away with the future does not need Washington to hobble its competitors.

Antitrust law, from the Sherman Act onward, exists to prevent the very outcome that increasing returns produce: a market owned by one or two firms. Here, that purpose is turned upside down. Washington is being lobbied to ban the one force that keeps AI markets from tipping to monopoly. No breakup is required, no decade-long case; the market has produced its own remedy, for free, and asks only that it not be thwarted. Block open models, and danger will not have been prevented. We will have manufactured the monopoly ourselves.

Adam Smith taught that competition disciplines even the mightiest firm, and W. Brian Arthur warned that with software the mighty can pull away before competition catches them. Open models are how the invisible hand answers Arthur with Smith — the Darwinian check that keeps the race alive. It would be a strange reading of American economic history for the government’s contribution to be shooting the solution.